Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

flux.1 dev Text Encode in very low cpu usage #6327

Open
AhBumm opened this issue Jan 3, 2025 · 3 comments
Open

flux.1 dev Text Encode in very low cpu usage #6327

AhBumm opened this issue Jan 3, 2025 · 3 comments
Labels
Potential Bug User is reporting a bug. This should be tested.

Comments

@AhBumm
Copy link

AhBumm commented Jan 3, 2025

Expected Behavior

Use all cores of cpu for text encode

Actual Behavior

img_v3_02i5_74c007d3-9f84-4718-91cd-78ecdafb491g
It only use 1 of cpu core for text encode.

Steps to Reproduce

System: Ubuntu Server 24.04.1 LTS
Linux Kernel: 6.8.0-49-generic
image
run in common t2i workflow with flux dev fp8 model, could reproduce the problem

Debug Logs

no error and warn in log

Other

No response

@AhBumm AhBumm added the Potential Bug User is reporting a bug. This should be tested. label Jan 3, 2025
@ltdrdata
Copy link
Collaborator

ltdrdata commented Jan 3, 2025

Did you run it in --cpu mode?

@AhBumm
Copy link
Author

AhBumm commented Jan 3, 2025

Did you run it in --cpu mode?

No.. the DiT part of model could run in cuda device correctly

@ltdrdata
Copy link
Collaborator

ltdrdata commented Jan 3, 2025

  1. First, it is normal to maximize the use of the GPU rather than the CPU for text encoding.
  2. Regarding CPU core imbalance, refer to the following issue.
    CPU cores are not fully utilized when using standalone build for Windows (torch) #5178

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Potential Bug User is reporting a bug. This should be tested.
Projects
None yet
Development

No branches or pull requests

2 participants